File size: 18,045 Bytes
f4456af b691127 2332381 f4456af b691127 f4456af b691127 22fe41e 3079197 b691127 f4456af 22fe41e f4456af 3079197 b691127 3079197 f4456af 3079197 028acc4 3079197 22fe41e f4456af 22fe41e 1eb186a b691127 f4456af b691127 f4456af b691127 22fe41e f4456af b691127 22fe41e f4456af b691127 f4456af b691127 f4456af 22fe41e b691127 f4456af b691127 22fe41e b691127 f4456af b691127 f4456af b691127 f4456af b691127 f4456af 22fe41e f4456af 22fe41e f4456af 22fe41e f4456af b691127 c372afe b691127 c372afe b691127 c372afe 22fe41e c372afe 22fe41e c372afe b691127 f4456af b691127 1eb186a b691127 1eb186a b691127 f4456af 22fe41e b691127 f4456af b691127 f4456af b691127 22fe41e b691127 f4456af b691127 f4456af b691127 22fe41e b691127 f4456af b691127 22fe41e b691127 f4456af b691127 f4456af 22fe41e b691127 f4456af b691127 f4456af b691127 f4456af b691127 f4456af b691127 f4456af b691127 f4456af b691127 3079197 b691127 22fe41e b691127 22fe41e 3079197 b691127 22fe41e b691127 22fe41e b691127 22fe41e b691127 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 |
import re
import json
import time
import os
from typing import List, Dict
import elasticsearch
import copy
from elasticsearch import Elasticsearch
from elasticsearch_dsl import UpdateByQuery, Q, Search, Index
from elastic_transport import ConnectionTimeout
from api.utils.log_utils import logger
from rag import settings
from rag.utils import singleton
from api.utils.file_utils import get_project_base_directory
import polars as pl
from rag.utils.doc_store_conn import DocStoreConnection, MatchExpr, OrderByExpr, MatchTextExpr, MatchDenseExpr, FusionExpr
from rag.nlp import is_english, rag_tokenizer
logger.info("Elasticsearch sdk version: "+str(elasticsearch.__version__))
@singleton
class ESConnection(DocStoreConnection):
def __init__(self):
self.info = {}
for _ in range(10):
try:
self.es = Elasticsearch(
settings.ES["hosts"].split(","),
basic_auth=(settings.ES["username"], settings.ES["password"]) if "username" in settings.ES and "password" in settings.ES else None,
verify_certs=False,
timeout=600
)
if self.es:
self.info = self.es.info()
logger.info("Connect to es.")
break
except Exception:
logger.exception("Fail to connect to es")
time.sleep(1)
if not self.es.ping():
raise Exception("Can't connect to ES cluster")
v = self.info.get("version", {"number": "5.6"})
v = v["number"].split(".")[0]
if int(v) < 8:
raise Exception(f"ES version must be greater than or equal to 8, current version: {v}")
fp_mapping = os.path.join(get_project_base_directory(), "conf", "mapping.json")
if not os.path.exists(fp_mapping):
raise Exception(f"Mapping file not found at {fp_mapping}")
self.mapping = json.load(open(fp_mapping, "r"))
"""
Database operations
"""
def dbType(self) -> str:
return "elasticsearch"
def health(self) -> dict:
return dict(self.es.cluster.health()) + {"type": "elasticsearch"}
"""
Table operations
"""
def createIdx(self, indexName: str, knowledgebaseId: str, vectorSize: int):
if self.indexExist(indexName, knowledgebaseId):
return True
try:
from elasticsearch.client import IndicesClient
return IndicesClient(self.es).create(index=indexName,
settings=self.mapping["settings"],
mappings=self.mapping["mappings"])
except Exception:
logger.exception("ES create index error %s" % (indexName))
def deleteIdx(self, indexName: str, knowledgebaseId: str):
try:
return self.es.indices.delete(indexName, allow_no_indices=True)
except Exception:
logger.exception("ES delete index error %s" % (indexName))
def indexExist(self, indexName: str, knowledgebaseId: str) -> bool:
s = Index(indexName, self.es)
for i in range(3):
try:
return s.exists()
except Exception as e:
logger.exception("ES indexExist")
if str(e).find("Timeout") > 0 or str(e).find("Conflict") > 0:
continue
return False
"""
CRUD operations
"""
def search(self, selectFields: list[str], highlightFields: list[str], condition: dict, matchExprs: list[MatchExpr], orderBy: OrderByExpr, offset: int, limit: int, indexNames: str|list[str], knowledgebaseIds: list[str]) -> list[dict] | pl.DataFrame:
"""
Refers to https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl.html
"""
if isinstance(indexNames, str):
indexNames = indexNames.split(",")
assert isinstance(indexNames, list) and len(indexNames) > 0
assert "_id" not in condition
s = Search()
bqry = None
vector_similarity_weight = 0.5
for m in matchExprs:
if isinstance(m, FusionExpr) and m.method=="weighted_sum" and "weights" in m.fusion_params:
assert len(matchExprs)==3 and isinstance(matchExprs[0], MatchTextExpr) and isinstance(matchExprs[1], MatchDenseExpr) and isinstance(matchExprs[2], FusionExpr)
weights = m.fusion_params["weights"]
vector_similarity_weight = float(weights.split(",")[1])
for m in matchExprs:
if isinstance(m, MatchTextExpr):
minimum_should_match = "0%"
if "minimum_should_match" in m.extra_options:
minimum_should_match = str(int(m.extra_options["minimum_should_match"] * 100)) + "%"
bqry = Q("bool",
must=Q("query_string", fields=m.fields,
type="best_fields", query=m.matching_text,
minimum_should_match = minimum_should_match,
boost=1),
boost = 1.0 - vector_similarity_weight,
)
if condition:
for k, v in condition.items():
if not isinstance(k, str) or not v:
continue
if isinstance(v, list):
bqry.filter.append(Q("terms", **{k: v}))
elif isinstance(v, str) or isinstance(v, int):
bqry.filter.append(Q("term", **{k: v}))
else:
raise Exception(f"Condition `{str(k)}={str(v)}` value type is {str(type(v))}, expected to be int, str or list.")
elif isinstance(m, MatchDenseExpr):
assert(bqry is not None)
similarity = 0.0
if "similarity" in m.extra_options:
similarity = m.extra_options["similarity"]
s = s.knn(m.vector_column_name,
m.topn,
m.topn * 2,
query_vector = list(m.embedding_data),
filter = bqry.to_dict(),
similarity = similarity,
)
if matchExprs:
s.query = bqry
for field in highlightFields:
s = s.highlight(field)
if orderBy:
orders = list()
for field, order in orderBy.fields:
order = "asc" if order == 0 else "desc"
orders.append({field: {"order": order, "unmapped_type": "float",
"mode": "avg", "numeric_type": "double"}})
s = s.sort(*orders)
if limit > 0:
s = s[offset:limit]
q = s.to_dict()
# logger.info("ESConnection.search [Q]: " + json.dumps(q))
for i in range(3):
try:
res = self.es.search(index=indexNames,
body=q,
timeout="600s",
# search_type="dfs_query_then_fetch",
track_total_hits=True,
_source=True)
if str(res.get("timed_out", "")).lower() == "true":
raise Exception("Es Timeout.")
logger.info("ESConnection.search res: " + str(res))
return res
except Exception as e:
logger.exception("ES search [Q]: " + str(q))
if str(e).find("Timeout") > 0:
continue
raise e
logger.error("ES search timeout for 3 times!")
raise Exception("ES search timeout.")
def get(self, chunkId: str, indexName: str, knowledgebaseIds: list[str]) -> dict | None:
for i in range(3):
try:
res = self.es.get(index=(indexName),
id=chunkId, source=True,)
if str(res.get("timed_out", "")).lower() == "true":
raise Exception("Es Timeout.")
if not res.get("found"):
return None
chunk = res["_source"]
chunk["id"] = chunkId
return chunk
except Exception as e:
logger.exception(f"ES get({chunkId}) got exception")
if str(e).find("Timeout") > 0:
continue
raise e
logger.error("ES search timeout for 3 times!")
raise Exception("ES search timeout.")
def insert(self, documents: list[dict], indexName: str, knowledgebaseId: str) -> list[str]:
# Refers to https://www.elastic.co/guide/en/elasticsearch/reference/current/docs-bulk.html
operations = []
for d in documents:
assert "_id" not in d
assert "id" in d
d_copy = copy.deepcopy(d)
meta_id = d_copy["id"]
del d_copy["id"]
operations.append(
{"index": {"_index": indexName, "_id": meta_id}})
operations.append(d_copy)
res = []
for _ in range(100):
try:
r = self.es.bulk(index=(indexName), operations=operations,
refresh=False, timeout="600s")
if re.search(r"False", str(r["errors"]), re.IGNORECASE):
return res
for item in r["items"]:
for action in ["create", "delete", "index", "update"]:
if action in item and "error" in item[action]:
res.append(str(item[action]["_id"]) + ":" + str(item[action]["error"]))
return res
except Exception as e:
logger.warning("Fail to bulk: " + str(e))
if re.search(r"(Timeout|time out)", str(e), re.IGNORECASE):
time.sleep(3)
continue
return res
def update(self, condition: dict, newValue: dict, indexName: str, knowledgebaseId: str) -> bool:
doc = copy.deepcopy(newValue)
del doc['id']
if "id" in condition and isinstance(condition["id"], str):
# update specific single document
chunkId = condition["id"]
for i in range(3):
try:
self.es.update(index=indexName, id=chunkId, doc=doc)
return True
except Exception as e:
logger.exception(f"ES failed to update(index={indexName}, id={id}, doc={json.dumps(condition, ensure_ascii=False)})")
if str(e).find("Timeout") > 0:
continue
else:
# update unspecific maybe-multiple documents
bqry = Q("bool")
for k, v in condition.items():
if not isinstance(k, str) or not v:
continue
if isinstance(v, list):
bqry.filter.append(Q("terms", **{k: v}))
elif isinstance(v, str) or isinstance(v, int):
bqry.filter.append(Q("term", **{k: v}))
else:
raise Exception(f"Condition `{str(k)}={str(v)}` value type is {str(type(v))}, expected to be int, str or list.")
scripts = []
for k, v in newValue.items():
if not isinstance(k, str) or not v:
continue
if isinstance(v, str):
scripts.append(f"ctx._source.{k} = '{v}'")
elif isinstance(v, int):
scripts.append(f"ctx._source.{k} = {v}")
else:
raise Exception(f"newValue `{str(k)}={str(v)}` value type is {str(type(v))}, expected to be int, str.")
ubq = UpdateByQuery(
index=indexName).using(
self.es).query(bqry)
ubq = ubq.script(source="; ".join(scripts))
ubq = ubq.params(refresh=True)
ubq = ubq.params(slices=5)
ubq = ubq.params(conflicts="proceed")
for i in range(3):
try:
_ = ubq.execute()
return True
except Exception as e:
logger.error("ES update exception: " + str(e) + "[Q]:" + str(bqry.to_dict()))
if str(e).find("Timeout") > 0 or str(e).find("Conflict") > 0:
continue
return False
def delete(self, condition: dict, indexName: str, knowledgebaseId: str) -> int:
qry = None
assert "_id" not in condition
if "id" in condition:
chunk_ids = condition["id"]
if not isinstance(chunk_ids, list):
chunk_ids = [chunk_ids]
qry = Q("ids", values=chunk_ids)
else:
qry = Q("bool")
for k, v in condition.items():
if isinstance(v, list):
qry.must.append(Q("terms", **{k: v}))
elif isinstance(v, str) or isinstance(v, int):
qry.must.append(Q("term", **{k: v}))
else:
raise Exception("Condition value must be int, str or list.")
logger.info("ESConnection.delete [Q]: " + json.dumps(qry.to_dict()))
for _ in range(10):
try:
res = self.es.delete_by_query(
index=indexName,
body = Search().query(qry).to_dict(),
refresh=True)
return res["deleted"]
except Exception as e:
logger.warning("Fail to delete: " + str(filter) + str(e))
if re.search(r"(Timeout|time out)", str(e), re.IGNORECASE):
time.sleep(3)
continue
if re.search(r"(not_found)", str(e), re.IGNORECASE):
return 0
return 0
"""
Helper functions for search result
"""
def getTotal(self, res):
if isinstance(res["hits"]["total"], type({})):
return res["hits"]["total"]["value"]
return res["hits"]["total"]
def getChunkIds(self, res):
return [d["_id"] for d in res["hits"]["hits"]]
def __getSource(self, res):
rr = []
for d in res["hits"]["hits"]:
d["_source"]["id"] = d["_id"]
d["_source"]["_score"] = d["_score"]
rr.append(d["_source"])
return rr
def getFields(self, res, fields: List[str]) -> Dict[str, dict]:
res_fields = {}
if not fields:
return {}
for d in self.__getSource(res):
m = {n: d.get(n) for n in fields if d.get(n) is not None}
for n, v in m.items():
if isinstance(v, list):
m[n] = v
continue
if not isinstance(v, str):
m[n] = str(m[n])
# if n.find("tks") > 0:
# m[n] = rmSpace(m[n])
if m:
res_fields[d["id"]] = m
return res_fields
def getHighlight(self, res, keywords: List[str], fieldnm: str):
ans = {}
for d in res["hits"]["hits"]:
hlts = d.get("highlight")
if not hlts:
continue
txt = "...".join([a for a in list(hlts.items())[0][1]])
if not is_english(txt.split(" ")):
ans[d["_id"]] = txt
continue
txt = d["_source"][fieldnm]
txt = re.sub(r"[\r\n]", " ", txt, flags=re.IGNORECASE|re.MULTILINE)
txts = []
for t in re.split(r"[.?!;\n]", txt):
for w in keywords:
t = re.sub(r"(^|[ .?/'\"\(\)!,:;-])(%s)([ .?/'\"\(\)!,:;-])"%re.escape(w), r"\1<em>\2</em>\3", t, flags=re.IGNORECASE|re.MULTILINE)
if not re.search(r"<em>[^<>]+</em>", t, flags=re.IGNORECASE|re.MULTILINE):
continue
txts.append(t)
ans[d["_id"]] = "...".join(txts) if txts else "...".join([a for a in list(hlts.items())[0][1]])
return ans
def getAggregation(self, res, fieldnm: str):
agg_field = "aggs_" + fieldnm
if "aggregations" not in res or agg_field not in res["aggregations"]:
return list()
bkts = res["aggregations"][agg_field]["buckets"]
return [(b["key"], b["doc_count"]) for b in bkts]
"""
SQL
"""
def sql(self, sql: str, fetch_size: int, format: str):
logger.info(f"ESConnection.sql get sql: {sql}")
sql = re.sub(r"[ `]+", " ", sql)
sql = sql.replace("%", "")
replaces = []
for r in re.finditer(r" ([a-z_]+_l?tks)( like | ?= ?)'([^']+)'", sql):
fld, v = r.group(1), r.group(3)
match = " MATCH({}, '{}', 'operator=OR;minimum_should_match=30%') ".format(
fld, rag_tokenizer.fine_grained_tokenize(rag_tokenizer.tokenize(v)))
replaces.append(
("{}{}'{}'".format(
r.group(1),
r.group(2),
r.group(3)),
match))
for p, r in replaces:
sql = sql.replace(p, r, 1)
logger.info(f"ESConnection.sql to es: {sql}")
for i in range(3):
try:
res = self.es.sql.query(body={"query": sql, "fetch_size": fetch_size}, format=format, request_timeout="2s")
return res
except ConnectionTimeout:
logger.exception("ESConnection.sql timeout [Q]: " + sql)
continue
except Exception:
logger.exception("ESConnection.sql got exception [Q]: " + sql)
return None
logger.error("ESConnection.sql timeout for 3 times!")
return None
|